Risk influencing factors in maritime accidents

126 Pages·2015·2.53 MB·English

Social Research
Per Stornes
P
e
r
S
t
o
r
n
e
s
Risk influencing factors
R
is
k
in in maritime accidents
fl
u
e
n
c
in
g
f
a
c
t
o
rs An exploratory statistical analysis of the
in
m Norwegian Maritime Authority incident
a
rit database
im
e
a
c
c
id
e
n
t
s
D7N4roa9rg1w vTaoryoll nadllhée 3im8 B Re
p
o
Tel: [phone] rt
Web: www.samforsk.no 20 Report 2015
1
5
Studio Apertura
Social Research

Per Stornes

Risk influencing factors
in maritime accidents

An exploratory statistical analysis of the
Norwegian Maritime Authority incident
database

Studio Apertura, NTNU Social Research
REPORT
     TITLE
Risk influencing factors in maritime accidents. An
NTNU Social Reseach
exploratory statistical analysis of the Norwegian
Studio Apertura Maritime Authority incident database.
Mailing Address:: NTNU Dragvoll, N-7491 Trondheim
Visiting address:: Dragvoll Allé 38B,
AUTHOR
Phone: [phone]
Per Stornes
Fax: (+47) 73 59 62 24
E-mail: [email protected]
FUNDER
Web.: www.samforsk.no
Norwegian Research Council, Norwegian Maritime
Business reg. number:. NO 986 243 836 Directorate, Norwegian Coastal Authority
REPORT NUMBER.. GRADATION FUNDER’S REF.
Årstall:Løpenr Open National Ship Risk Model
ISBN PROJECT NUMBER. NUMBER OF PAGES
978-82-7570-427-4 (trykk) / 978-82-7570-428-1 (web) 2605 124
PRICE (excl. postage and handling) PROJECT MANAGER QUALITY ASSURED BY
Petter Almklov Trond Kongsvik
DATO APPROVED BY:
May 18, 2015 Petter Almklov
ABSTRACT
This report is an exploratory statistical analysis of the Norwegian Maritime Authority’s database, using data
on groundings, collisions, allisions fires/explosions and some data on capsizings in Norwegian waters from
1981 through 2014. The analysis is part of the Norwegian Ship Risk Model project. The statistical analysis is
divided into two parts.
The first part is a descriptive analysis, which identifies common traits in accidents. Vessel types are broken
down into 12 categories. Fires/explosions are most common on small fishing vessels in outer coastal waters in
Northern regions, with a notable proportion happening in dock. Groundings are most common on cargo vessels
in narrow coastal waters in Northern regions, and in the dark. Capsizings typically involve small fishing and
cargo vessels in outer coastal waters, and feature strong winds and higher seas. Collisions are most frequent
among fishing and break bulk vessels in outer coastal waters. Allisions are common among medium sized
passenger vessels in the harbour area.
The second part of the analysis uses multinomic regression to describe variation between accidents. Accident
types vary modestly between vessel types, gross tonnages and length, but substantially between waters.
Groundings are associated with narrow coastal waters, collisions with outer coastal, allisions with port areas,
and fires/explosions are associated with vessels in dock. Weather has limited effect, although collisions are ten
times more likely than other accidents under no visibility. High vessel damage severity was primarily associated
with shorter vessels. We were not able to explain much variation in injuries and fatalities.
The results indicate a strong need to connect accident data with normalized traffic data to identify risk
influencing factors with more certainty.

Risk influencing factors, maritime accidents, relative probabilities, multinomic
KEYWORDS
regression, fires/explosions, groundings, collisions, allisions, vessel types,
waters, visibility, common accident scenario.

ii
1 Preface
This report was written in January-March 2015 on commission from Studio Apertura,
NTNU Social Research, as part of the National Ship Risk Model project. The author would
like to thank the entire staff at Studio Apertura for helping me out with all my needs in this
period. In particular, I would like to thank Trond Kongsvik and Petter Almklov for valuable
comments and support in the work on this report. In addition, I would like to thank Rolf Bye,
Stein Haugen and Elisabeth Blix at Safetec Trondheim for their input in the research process.
Special thanks goes to the Norwegian Maritime Authority, and their representative Vegar
Berntsen in particular, for assistance in supplying data and information.

Trondheim, May 18, 2015,

Per Stornes
iii
2 Contents
1 Preface .......................................................................................................................... iii 
2 Contents ........................................................................................................................ iv 
3 Abstract ......................................................................................................................... vi 
4 Introduction .................................................................................................................... 8 
4.1 The «National ship risk model» project .................................................................. 8 
4.2 Goals of Work Package 3. ....................................................................................... 8 
4.3 Layout of the report. ................................................................................................ 9 
5 Theoretical perspectives .............................................................................................. 10 
5.1 Risk models ........................................................................................................... 12 
5.2 Research questions ................................................................................................ 13 
6 The NMA incident database. ....................................................................................... 14 
6.1 History of the database. ......................................................................................... 14 
6.2 Validity of the data ................................................................................................ 15 
6.3 Accident types. ...................................................................................................... 16 
6.4 Other qualities of accidents ................................................................................... 17 
6.5 Vessel groups in the database ............................................................................... 18 
6.6 Vessel types in the database. ................................................................................. 19 
6.7 Vessel categories in the present analysis. ............................................................. 20 
6.8 Vessel properties ................................................................................................... 22 
6.9 Geographical properties ........................................................................................ 24 
6.10 Weather properties .............................................................................................. 26 
6.11 Date and time properties ..................................................................................... 27 
6.12 Certification properties........................................................................................ 27 
6.13 Vessel identity ..................................................................................................... 29 
6.14 A note on missing data ........................................................................................ 30 
7 Analysis Part 1: Common traits in accidents - descriptive statistics. ......................... 31 
7.1 Vessel qualities...................................................................................................... 31 
7.2 Geographical qualities ........................................................................................... 34 
7.3 Weather qualities. .................................................................................................. 35 
7.4 Time qualities ........................................................................................................ 36 
7.5 Other notable qualities .......................................................................................... 38 
8 Analysis Part 1: Common traits in accidents ............................................................... 40 
8.1 Common traits in fires and explosions .................................................................. 40 
8.2 Common traits in groundings ................................................................................ 41 
8.3 Common traits in capsizings ................................................................................. 42 
8.4 Common traits in collisions .................................................................................. 43 
8.5 Common traits in allisions .................................................................................... 43 
8.6 Common traits in accidents: A summary and comparison. ................................... 44 
8.7 Risk influencing factors based on common traits ................................................. 46 
9 Analysis part 2: Logistic and multinomic regression methods .................................... 47 
9.1 The logistic regression model. .............................................................................. 47 
9.2 The multinomic regression model ......................................................................... 51 

iv
10 Analysis part 2: Regression analyses of accidents. ................................................... 56 
10.1 Preliminary analysis of vessel types and accidents. ............................................ 56 
10.2 Multinomic analysis of vessels, geography and weather. ................................... 57 
10.3 Integrated model of vessel, geography and weather. .......................................... 63 
10.4 Predicted probabilities from the integrated model .............................................. 68 
10.5 Conditional probabilities ..................................................................................... 70 
10.6 Predicted probabilities of common traits ............................................................ 78 
10.7 High risk profiles of accidents ............................................................................ 80 
10.8 Multinomic analysis of time categories. ............................................................. 85 
10.9 Multinomic analysis of certification. .................................................................. 87 
10.10 Multinomic analysis of operational state. ......................................................... 88 
10.11 Logistic regression analysis of severity ............................................................ 90 
10.12 Logistic regression analysis of injuries ............................................................. 95 
10.13 Logistic regression analysis of fatalities. .......................................................... 98 
11 Discussion. ............................................................................................................... 101 
11.1 Vessels, geography, weather and maritime accidents. ...................................... 101 
11.2 Time, certification and operational state in maritime accidents. ...................... 102 
11.3 Damage severity, injuries and fatalities ............................................................ 103 
12 Conclusion: Risk influencing factors in maritime accidents. .................................. 104 
13 Literature .................................................................................................................. 105 
14 Appendix: Descriptive statistics for regression analysis ......................................... 107 
14.1 Vessel qualities.................................................................................................. 107 
14.2 Geographical qualities ....................................................................................... 110 
14.3 Weather qualities. .............................................................................................. 110 
14.4 Time qualities .................................................................................................... 111 
14.5 Other notable qualities ...................................................................................... 113 
14.6 Appendix: Correspondence analysis of accidents and vessels. ......................... 115 
14.7 Appendix: NMA vessel codes. .......................................................................... 118 
14.8 Appendix: Cargo vessel types with translations. .............................................. 119 
14.9 Appendix: Map of Norwegian waters ............................................................... 122 

v
3 Abstract
This research report is an exploratory statistical analysis of the Norwegian Maritime
Authority’s (NMA) incident database, with the objective of exploring possible Risk Influencing
Factors (RIFs) in Norwegian maritime traffic for the National Ship Risk Model project. I use data
on vessel accident reported from 1981 through 2014 on groundings, collisions, allisions and
fires/explosions. Capsizings are also included in part 1 of the analysis. The analysis divides
vessels into 12 categories.
The analysis consists of two main parts. The first part is a descriptive analysis where I
describe common traits of accidents.
Fires and explosions are most common on small fishing vessels in outer coastal waters in
the Northern regions. They usually happen in good weather. A notable proportion of fires and
explosions happen while the vessel is in dock.
Groundings are most common among cargo vessels, however small fishing vessels in
coastal fishing are also notable. Narrow coastal waters are typical, as is the northernmost region of
the coastline. Groundings are most common in the dark and at night, while the ship is underway.
Capsizings typically involve smaller fishing and cargo vessels in outer coastal waters. The
northernmost region of the coastline is once again notable for capsizings. Capsizings are
characterized by strong winds, and more frequent in moderate and high seas than other types of
accidents.
Collisions are most common among fishing vessels and break bulk vessels. They are most
frequent in outer coastal waters, but narrow coastal waters and harbour areas also feature notably
in collisions. Once again, the northernmost coastal region reports most collisions. Collisions tend
to happen in good weather conditions.
Allisions are most common among medium sized passenger vessels, in particular ferries.
They tend to feature vessels certified for trafficking protected waters. Half of all allisions are
reported in narrow coastal waters, which in practice usually means striking the quay. Most
allisions happen in the two regions between Lindesnes and Trondheim. They tend to happen in
good weather, but a larger proportion happens in in strong winds compared to other accidents.
The second part of the analysis is an advanced statistical analysis. I perform a multinomic
regression on accidents, and compare the relative influence on vessel types and qualities,
geographical qualities and weather qualities in an integrated model. In addition, I perform
multinomic analyses of certification, operational states and time variation. I perform logistic
regressions on damage severity, injuries and fatalities using variables from the integrated model.
In addition, I predict conditional probabilities for results from these analyses.
vi
The main results are as follows. For vessel types, I find significant but modest differences
in accident qualities for ferries, passenger/cruise vessels, high speed craft, work & service vessels
and break bulk vessels. Foreign vessels are more likely to experience groundings than other
accidents. Higher gross tonnages are associated with decreased probabilities of groundings.
Accidents vary substantially between waters. Groundings are most likely in narrow coastal
waters, collisions most likely in outer coastal waters, and allisions most likely in port areas.
Fires/explosions are more probable along quay than in other waters.
Weather has a limited effect on accident probabilities. Collisions are ten times more likely
under conditions of no visibility.
Variations in time were modest. Groundings are more likely by night than by day, and
collisions less likely. Vessels certified for coastal fishing had the highest probability of
fires/explosions. Allisions appear more likely on arrival of port than on departure.
Vessel damage severity was primarily associated with vessel length. The shorter the vessel,
the higher the odds of severe damage, particularly in allisions. I were not able to explain much of
the variation on injuries and fatalities. High seas increase the risk of injury substantially in
fires/explosions, whereas high speed craft have around five times higher probability of injuries
than large fishing vessels. Groundings in short vessels increase the probability of fatalities by over
30 times.
I propose the following main risk influencing factors.
For fires/explosions, fishing vessels appear at high risk. Large gross tonnages increase risk
of fires, as well as longer vessels. The risk of fires is high at the quayside, while weather appears
to be little influential.
For groundings, cargo vessels (work and service vessels in particular) appear at higher risk.
Vessels of low gross tonnage and longer length appear at higher risk. Narrow coastal waters
increase the relative risk of groundings substantially.
For collisions, small break bulk vessels appear at higher risk. Travelling in no visibility
increases the relative risk of a collision considerably. Outer coastal waters increases the risk of a
collision considerably.
For allisions, high speed craft of medium gross tonnage and longer lengths appear at higher
risk of allisions. Allisions are closely tied to the harbour area.

vii
4 Introduction
This research report is an exploratory statistical analysis of the Norwegian Maritime
Authority’s (NMA) incident database, with the objective of identifying Risk Influencing
Factors (RIFs) in Norwegian maritime traffic. The following chapter gives a brief description
of the National Ship Risk Model project, the objectives of the second work package in the
project, and an overview of the contents of this report.
4.1 The «National ship risk model» project
The National Ship Risk Model (NSRM) is a joint research project with the ultimate
objective of developing a risk model for traffic in Norwegian waters. The research group
consists of Studio Apertura at NTNU Social Research, Safetec Nordic AS and NTNU. The
project is funded by the Norwegian Maritime Authority, the Norwegian Coastal
Administration and the Norwegian Research Council (NTNU Social Research 2014).
The NSRM will be used to better monitor and communicate the risk picture of maritime
activities in Norwegian waters. It will be used by the NMA to monitor changes in the risk
picture, prioritize inspection activities (risk based inspections), and support decisions
regarding development of regulations and safety improving measures. Furthermore, the risk
model will be used as a tool by the NCA to improve the quality of their risk analysis
preceding major interventions and modifications of fairways and ports, as well as in the daily
risk assessments performed by the Vessel Traffic Service (VTS) Centres. It will also be used
in the decision processes related to the pre-deployment of the governmental tugboat
contingency service as well as oil spill response measures.
4.2 Goals of Work Package 3.
The objective of this work package is to generate knowledge regarding causes and
conditional factors associated with different types of marine accidents. This knowledge will
be generated by conducting statistical exploratory analysis of accident data. The data will be
analysed by using explorative methods of logistic and multinomic regression analysis. The
dependent variables will be accident types and accident qualities such as damage severity,
injuries and fatalities. Parameters for the independent variables in the explorative analysis will
be qualities of vessels, geography and weather. The results of these analyses will improve the
knowledge regarding causes of marine accidents. The conceptualization of RIFs is based on
the assumption that the risk (in terms of a quantitative measure) can be controlled by
changing/managing/controlling the Risk influencing factors. The identification of RIFs will
8